The argument centers around the fallibility of some (naive) formal proofs of Friendliness which I’ve seen people discussing the AI box problem willing to accept.
This post might make more sense to me if you presented these proofs.
This post might make more sense to me if you presented these proofs.